Research And Realization Of Speech Noise Suppression Technology In Airborne Environments

Speech is one of the main intercommunion tools between persons or between person and machinery. In the domain of speech signals processing, it's a significant research that how to implement speech's transmission and storage effectively. Speech noise suppression is an important technology of speech signal processing. For many years many progresses have been made in this field to find good methods for it.The mostly important purpose of noise suppression is to improve speech quality, enhance speech intelligibility and reduce the listener's ear fatigue at the same time. Firstly, Voice activity detection methods were introduced. The detecting performance was analyzed and compared between some algorithms. And then the dissertation mainly ameliorated the algorithm based on auto-correlation function. It's proved in emulation that the improved algorithm is more effective than traditional algorithm based on auto-correlation.Speech enhancement is the most important part of speech noise reduction. Most applied speech enhancement algorithms are realized by precise methods of noise estimation and favorable enhancement algorithms. So in second part a effective method for noise estimation was introduced. And then the thesis introduced traditional spectral subtraction algorithm and other two speech enhancement algorithms which based on Minimum Mean Square Error (MMSE) and associated with auditory masking properties. It's proved in emulation that traditional spectral subtraction algorithm and the algorithm based on MMSE were not effective in aviation circumstance. Although the effect of the method based on auditory masking properties was better, its practicability was restricted by its complexity. So in the dissertation an algorithm was adopted which combined effective noise estimation and traditional spectral subtraction algorithm. The test proved that this method could track the characteristics of aviation noise favorably and its operation efficiency was higher than others'.At last, according to a new relative measurement of speech intelligibility and taking into advantages and disadvantages of all algorithms account, this paper selected a speech enhancement algorithm suitable for aviation noise background. At the same time in order to improve the listener's hearing fatigue, an approach of comfortable noise generation was presented. Based on all of these analyses, a whole noise suppression flow in aviation circumstance was presented and its validity and applicability were proved both in Matlab and on hardware flat.